[patch 5/9] unprivileged mounts: allow unprivileged bind mounts

Karel Zak kzak at redhat.com
Wed Jan 9 05:26:10 PST 2008


On Wed, Jan 09, 2008 at 01:45:09PM +0100, Jan Engelhardt wrote:
> 
> On Jan 8 2008 20:08, Miklos Szeredi wrote:
> >> On Tue, 2008-01-08 at 12:35 +0100, Miklos Szeredi wrote:
> >> > +static int reserve_user_mount(void)
> >> > +{
> >> > +       int err = 0;
> >> > +
> >> > +       spin_lock(&vfsmount_lock);
> >> > +       if (nr_user_mounts >= max_user_mounts && !capable(CAP_SYS_ADMIN))
> >> > +               err = -EPERM;
> >> > +       else
> >> > +               nr_user_mounts++;
> >> > +       spin_unlock(&vfsmount_lock);
> >> > +       return err;
> >> > +} 
> >> 
> >> Would -ENOSPC or -ENOMEM be a more descriptive error here?  
> >
> >The logic behind EPERM, is that this failure is only for unprivileged
> >callers.  ENOMEM is too specifically about OOM.  It could be changed
> >to ENOSPC, ENFILE, EMFILE, or it could remain EPERM.  What do others
> >think?
> 
> ENOSPC: No space remaining on device => 'wth'.
> ENOMEM: I usually think of a userspace OOM (e.g. malloc'ed out all of your
> 32-bit address space on 32-bit processes)
> EMFILE: "Too many open files"
> ENFILE: "Too many open files in system".
> 
> ENFILE seems like a temporary winner among these four.

 I see "EMFILE", it's still supported by the latest mount(8).

> Back in the old days, when the number of mounts was limited in Linux,
> what error value did it return? That one could be used.

 Copy & past from mount-0.99.2:

  /* Mount failed, complain, but don't die.  */
  switch (mnt_err)
    {
    case EPERM:
      if (geteuid() == 0)
	error ("mount: mount point %s is not a directory", node);
      else
	error ("mount: must be superuser to use mount");
      break;
    case EBUSY:
      error ("mount: wrong fs type, %s already mounted, %s busy, "
	"or other error", spec, node);
      break;
    case ENOENT:
      error ("mount: mount point %s does not exist", node); break;
    case ENOTDIR:
      error ("mount: mount point %s is not a directory", node); break;
    case EINVAL:
      error ("mount: %s not a mount point", spec); break;
    case EMFILE:
      error ("mount table full"); break;
    case EIO:
      error ("mount: %s: can't read superblock", spec); break;
    case ENODEV:
      error ("mount: fs type %s not supported by kernel", type); break;
    case ENOTBLK:
      error ("mount: %s is not a block device", spec); break;
    case ENXIO:
      error ("mount: %s is not a valid block device", spec); break;
    case EACCES:
      error ("mount: block device %s is not permitted on its filesystem", spec);
      break;
    default:
      error ("mount: %s", strerror (mnt_err)); break;
    }


  Karel

-- 
 Karel Zak  <kzak at redhat.com>


More information about the Containers mailing list